Frequency based adaptive spatial resolution selection for 3D video coding

Downsampling applied to texture views prior to encoding can increase the subjective quality of decoded video. In our study, we show that spatial resolution selection based on traditional pixel-based distortion metrics, such as Mean Square Error (MSE) is weakly correlated with the resolution selection based on subjective quality of coded video. To overcome this problem, we propose a novel frequency-based distortion metric which is shown to resemble subjective quality of coded video more accurately compared to conventionally used MSE-based metric.